Features
Everything Trippi Cue does while the interview runs.
One job, done in one place: notice the question that was aimed at you, put a short answer in front of you before the moment passes, and keep the conversation readable in your own language. Here is the whole of it, with nothing added that the extension cannot do.
One session, end to end
From your click to the card on screen.
Five steps, and three of them exist to keep the window quiet. Most of an interview is somebody else talking, and a copilot that forgets that becomes noise you close halfway through.
-
1
You click the icon
Chrome hands a tab’s sound to an extension only after a real click on it. That click is the whole switch: no autostart, no background listening, nothing to forget to turn off.
-
2
It hears the call
The audio of that one tab goes to speech recognition and comes back as lines with speaker names. Only the last dozen lines are held in the tab while the call runs.
-
3
A line looks like a request
A cheap local check throws away the “yeah”, “sure” and “one second” of a conversation. What is left — a question, an instruction, a line with your name in it — is worth asking a model about.
-
4
The model decides who was asked
It sees the recent lines, the participants and your own name, and answers one question: was this addressed to you? On a panel that is the whole job, and if the answer is no, nothing is shown.
-
5
A card appears, translated underneath
The question as it was heard, a short answer you can say in your own words, and the line itself in your language. Below the confidence threshold, or the same question twice in a minute, and no card appears at all.
What you get
Six things, and the first one is the point.
Live answers
Two or three sentences, written to be said out loud rather than read from a screen. The answer is in the language you pick; the question stays in the language it was asked in.
Question detection
Questions, follow-ups and instructions — “walk me through”, “what would you do if” — are caught without a keyword or a hotkey. The threshold and the dedup window are ours to tune, and they are why the window stays quiet most of the call.
“About this meeting”
One plain-text line before the call — the role, the round, the two stories you want to tell — and every answer is written against it instead of against nobody in particular.
Ask it yourself
A field at the bottom of the window sends your own question with the call as context. It skips every automatic gate — you asked, so it answers.
The call in your language
Every recognised line also arrives translated, from the first minute, and the answer language is a separate choice: 68 to pick from, switchable mid-call.
A window, not an overlay you fight
It moves, it resizes, it lives in its own shadow tree so the call’s styles cannot reach it — and its own styles cannot reach the call. Close it and the session pauses.
Before, during, after
The whole product lives inside the call.
Most meeting tools do their work after everyone has hung up. This one is useless afterwards and useful in the thirty seconds you are being looked at.
Before the call
One line in “About this meeting”: what this call is, what you are interviewing for, what you would rather talk about. Nothing to upload, nobody to invite, no bot to schedule.
During the call
Click once. Questions aimed at you turn into cards with short answers; anything else stays out of the way. The conversation runs underneath in your language, and you can ask your own question at any point.
After the call
Nothing is produced. No notes, no summary, no recording, no email an hour later — you close the window and it is over. The recognised text and the answers are kept for 30 days so we can tell a poor answer from a poor question, and are then deleted.
Honest about the gaps
What it does not do.
Said here rather than discovered on your third call. If one of these is what you actually need, we would rather you knew now.
It does not read your screen
A shared diagram, a coding pad, a slide — none of it reaches Cue. It works from what is said out loud, which in an interview is most of the round but never all of it.
No answers from your documents
Cue answers from what was said in the call and from your one line of context. There is nowhere to upload a CV or a portfolio, because nothing would read it.
No practice mode, no feedback
This is not an interview trainer. There are no mock rounds, no score afterwards and no advice on how you did — it works during the call and produces nothing after it.
No desktop apps
The Zoom and Teams desktop applications live outside the browser, where no extension can reach them. Cue works in browser tabs, on a computer.
Missing something on this list? Tell us which one — what people ask for is how the next thing gets built.
Platforms
Where all of this works.
Browser meetings in Chrome and Chromium browsers. Each platform has its own quirks, its own permission and its own troubleshooting page.
Privacy
Where the sound and the text go.
The audio of the tab is streamed from your browser straight to our speech-to-text provider over an encrypted connection — it does not pass through our servers and is never recorded. The recognised lines go to our backend and become one prompt for the model. That window and the answer are kept for 30 days, so a poor answer can be told from a poor question, and are then deleted automatically.
What we keep
- An install record — an anonymous identifier for your copy of the extension.
- Counters — minutes used and answers shown, by day. This is what your plan is measured against.
- An account, if you make one — your email, so a plan bought on one machine works on another.
- The call window and the answer, for 30 days — then deleted, never used to train a model, and gone with your account if you delete it.
No audio, ever: it does not reach us at all, and no recording or file of your call is produced at any point.
FAQ
Questions about the features.
How many cards can be on screen at once?+
Three. A fourth pushes out the oldest one. The window is meant to be glanced at mid-sentence, not read like a feed.
Why did it stay silent when I was asked something?+
Either the model judged the question was not aimed at you, or its confidence was below the threshold, or the same question already had a card in the last minute. You can always ask it yourself in the field at the bottom.
Can I turn off the automatic cards?+
Yes. With automatic answers off, the window only reacts to questions you type in yourself. The switch works when the answer arrives, not by asking the model nicely.
Does it answer questions asked of other people?+
No. That is the one thing the model is asked to decide, and the check runs again on our side when the answer comes back.
What happens when my minutes run out?+
The cards stop and the window says so. The interview is not interrupted in any way — Cue is a passenger in the tab, not part of the call.
Does it work if the interview is in another language?+
Yes, and the lines arrive translated as well as answered. Recognition follows the call, the answer follows your pick — 68 languages in the window.
One click, and the next question is not yours alone.
Trippi Cue is in the Chrome Web Store and starting is free. Add it, click the icon in your next interview and see what it catches.